Danksharding and cheaper data for rollups
Since rollups depend on posting data back to L1, the cost of that data is the main lever left for scaling the whole ecosystem further. Ethereum's answer is proto-danksharding and its planned successor, full danksharding, which introduce blob space, a dedicated, separately-priced data lane attached to blocks specifically for rollup data. Blobs are pruned after a fixed window rather than stored forever, since rollups only need the data available long enough for anyone to verify and challenge it, not permanently on every node. This keeps rollup data costs decoupled from the fee pressure of regular execution gas, which is what actually drove L2 fees down sharply when it shipped.
Full danksharding is meant to scale blob capacity much further by spreading the job of storing and verifying blob data across the validator set using data availability sampling, where individual nodes verify small random samples rather than every node storing everything. It's a genuinely difficult distributed systems problem, and describing it as fully solved would be overselling where the research and engineering actually stand today, meaningful pieces are live, and the full scale of the design is still being built out in stages.